April Update: Research-Driven Prompt Strategy & Midjourney v7 Improvements

Updated: April 7, 2025

I'm thrilled to announce our latest Midjourney Prompt Manager update! Since Midjourney v7 alpha launched late last week, I've conducted in-depth research and A/B testing with ChatGPt o1 pro to see how different prompt structures (short keyword lists vs. flowing descriptive sentences) impact results in Midjourney v7. Below, I’ll share my findings, outline my new strategy for Midjourney v7 prompt generation, and show you examples from four different tests.

1. Research & Testing

I embarked on a series of experiments to compare prompts:

Original, unstructured keyword lists – base prompt with simple, comma-separated keywords (minimal context).
New descriptive approach – weaving user inputs into a short, coherent sentence or two after the base prompt (followed by parameters).

I tested each approach against Midjourney v7 (and sometimes v6.1) using the exact same parameters like --stylize and --chaos (note that v7 doesn't allow --weird). My primary goals were realism, visual consistency, clarity of style and an overall final pleasing image.

Key Observations

Flowing Descriptions sometimes produce more diverse color palettes and smoother transitions.
Keyword-Style Prompts can yield bold, uniform styling or color emphasis, especially at higher --stylize or --chaos.
No Single Format is always best. Each approach might be more suitable depending on your creative vision. However, in general, I found that flowing descriptions produced more desirable results.

2. The New Strategy

After poring over test images and ChatGPT o1 pro feedback, I’ve updated the prompt creation logic inside Midjourney Prompt Manager slightly to enable more flowing descriptions:

Start with the user’s base prompt (the most important piece).
Merge all style selections (theme, mood, camera, colors, lighting) into one descriptive phrase.
- Example:
```
in a futuristic, cinematic style, viewed from a top-down angle, shot on IMAX...
```
Append any special Midjourney parameters at the end (e.g., --v 7 --ar 1:1), ensuring no duplicates.
Avoid over-labeling (like “Theme: …” or “Mood: …”) which can sometimes dilute emphasis (another prompt strategy I tested).

This approach still clarifies user choices but presents them to Midjourney in a succinct, natural way that helps speed up prompt creation and future iteration to maintain image consistency. This is particularly important now that Midjourney v7 does not support srefs.

3. Example Prompt

Here’s a sample Midjourney Prompt Manager output I used in one test (notice it's more natural language throughout compared to older prompts the app produced prior to the v7 alpha launch):

A hidden valley inhabited by ethereal creatures. The environment has a whimsical, otherworldly ambiance, with vibrant alien flora. Soft morning light at dawn casts subtle glimmers, while a surreal dreamlike aura envelops the entire scene in a magical realism, ethereal style, viewed from a low angle shot, shot on canon eos 5d mark iv, dawn lighting, at sunrise, with a fantasy vibe, accented by holographic glow, god rays, featuring vibrant colors, polychromatic, emerald green, gold, mint, teal, pastel colors, with flowing fabric, translucent, chromatic aberration, bioluminescence, phosphorescence effects -–ar 16:9 -–v 7 -–s 500 -–c 30

In short, it’s the same attributes as the short comma-separated list of keywords that was originally used, but rephrased into a coherent sentence. Let's take a look at some examples.

4. Test Results

To illustrate, I’ll show four sets of images where I applied original vs. new strategies. Each set contains two images: (a) and (b).

Test 1

Results:

1a (original):

The output here feels almost surreal—like a cinematic dream of a cyberpunk city. The stacked keywords provided strong style direction, but the result leans toward heightened saturation and stylized drama. You’ll notice bold neon lighting, slightly softened edges, and a surreal quality to the buildings and reflections, creating an almost painted or speculative world. The atmosphere is unified and visually striking, but the color transitions are more abstract than realistic.

1b (new):

This version introduces a sense of realism that the keyword-driven prompt didn’t quite capture. The lighting is more grounded, especially in the reflections and signage, which feel believable without being flat. You can see more environmental coherence—shadows align, structure depth is emphasized, and the perspective feels immersive. The prompt phrased in natural language gave Midjourney clearer contextual intent, and the result is more photographic and true-to-world, without losing style.

Test 2

Results:

2a (original):

The original prompt yields a confident, almost artificial boldness. Colors like pink and lavender pop clearly against darker greens, with strong outlines and a uniform vibrancy throughout the scene. The lighting feels direct and staged, offering a crisp, poster-like visual where the flowers almost seem rendered or posed. It’s eye-catching and clean—but the color transitions are abrupt, and the overall palette is more limited.

2b (new):

This version feels more immersive, as though you’ve stumbled into a naturally lit patch of enchanted forest. The lighting behaves more organically, with shallow depth of field and gentle bokeh in the foreground and background. Rain particles or dew glints introduce a soft shimmer. Colors transition more smoothly, and the tonal range—while still warm—includes more natural greens, golds, and shadows, giving the entire scene a richer, more layered emotional tone.

Test 3

Results:

3a (original):

The scene in 3a feels bold and stylized, with rich amber and pink hues dominating the crystal surfaces. There’s a clear visual punch, but the effect leans toward intensity over subtlety. Color transitions are sharp and deliberate, emphasizing vibrant reflections and uniform symmetry. The composition is almost editorial—gorgeous, but a bit flatter in mood. It reflects how strong keywords can anchor a visual idea with authority, but sometimes at the cost of nuance.

3b (new):

In contrast, 3b reads more like a painting caught mid-breath. The crystals shift fluidly from peach to lavender to soft mint green, with the surrounding water and sky responding in kind. Light rolls across the surfaces more gently, giving a luminous quality to both the highlights and shadows. The natural phrasing of the prompt allowed Midjourney to interpret the environment as a whole system—resulting in a more immersive, emotionally resonant image.

Test 4

For our fourth test, we focused on a futuristic runway scene featuring a model and a cyberpunk city backdrop. As usual, (a) is the original keyword list approach, while (b) is the new descriptive approach.

Test 4b - New Descriptive Approach — Image: 4b (new)

Results:

4a (original):

This image feels stylized and on-brand for a cyberpunk aesthetic, using short, directive keywords like “model,” “futuristic rooftop,” and “neon glow.” The resulting output is clean and literal: a futuristic outfit, a recognizable model figure, and a moody city background. The lighting is dramatic but somewhat uniform, and the subject feels isolated in frame. While the image is clear and confident, it lacks secondary context or storytelling cues beyond the obvious style tags.

4b (new):

This version feels more cinematic. The model still takes center stage, but now there’s a layered sense of environment and narrative—like a frame out of a sci-fi film. The lighting transitions more subtly across the outfit, and secondary elements (like additional figures and environmental reflections) help round out the scene. The use of natural phrasing in the prompt opened the door for Midjourney to interpret not just what is shown, but how the story is told visually—through light, lens choice, and character relationships.

5. Conclusion

Prompt Variation Matters: The smallest tweak—whether rearranging words or removing extra labels—can shift Midjourney’s interpretation.
Hybrid Approaches: Sometimes mixing short, direct style keywords into a semi-descriptive phrase yields the strongest synergy.
Ongoing Refinement: I plan to keep testing with upcoming Midjourney updates and I have made adjustments to the way Midjourney Prompt Manager creates high-quality prompts for Midjourney input. The user interface will likely evolve further to let you choose how verbose or minimal you want your final prompt phrasing.

Next Steps: Try it for yourself!

Daily, I share prompt instructions on X that enable easy inputs into Midjourney Prompt Manager and the format looks like this:

1. Start with a base prompt

A futuristic cityscape filled with flying cars. The environment has a cinematic and mysterious flair, showcasing bold street art. Overcast skies during twilight lend a dramatic atmosphere, while neon glows and reflections emphasize the advanced cyberpunk aesthetic.

2. Choose your Midjourney Parameters

‣ Aspect Ratio: 1:1
‣ Mode: raw
‣ Version: 7
‣ Quality: 1
‣ Stylize: 750
‣ Weirdness: 0 (not compatible with v7)
‣ Variety / Chaos (%): 50

3. Choose your Midjourney Prompt Manager Style Selections

‣ Theme: Cyberpunk
‣ Emotion & Mood: Mysterious
‣ Camera Angle & Shot Types: Bird’s Eye View Shot
‣ Cameras & Lenses: IMAX Digital Camera with 50mm anamorphic lens
‣ Natural Lighting: Overcast
‣ Time of the Day: Twilight
‣ Ambiance & Styling: Street Art
‣ Ambient Lighting: Neon Glow
‣ Lighting Effects: Specular Highlights
‣ Color Treatment: Vibrant Colors
‣ Number of Colors: Octachromatic
‣ Specific Colors: Deep Pink, Electric Yellow, Fiery Orange, Neon Purple, Silver
‣ Color Palettes: Twilight Scheme
‣ Physical Properties & States of Matter: Metallic
‣ Reflections & Refractions: Iridescent
‣ Chromism & Luminescence: Electroluminescence, Metallochromism
‣ Artist Style: (None specified)

Follow this workflow starting with this or your own base prompt, select the Midjourney Parameters, and go through the Midjourney Prompt Manager Style Categories and choose the keywords you want to guide the overall look of your image. 1, 2, and 3. Then copy the Key Prompt and input it into Midjourney for rendering. That's it!

Visit midjourneypromptmanager.com today to try out the new prompt strategy for yourself!

Looking Ahead

I remain committed to providing the best user experience for Midjourney v7 (and beyond). Keep an eye out for the upcoming beta launch this month! I'm adding the ability to save prompts and images for easy organization and iteration of your content and campaign images.

In the meantime, let us know your thoughts:
feedback@keypromptninja.io

Happy prompting!